To determine whether the social environment surrounding lesbian, gay, and bisexual youth may contribute to their higher rates of suicide attempts, controlling for individual-level risk factors.
A total of 31 852 11th grade students (1413 [4.4%] lesbian, gay, and bisexual individuals) in Oregon completed the Oregon Healthy Teens survey in 2006–2008. We created a composite index of the social environment in 34 counties, including (1) the proportion of same-sex couples, (2) the proportion of registered Democrats, (3) the presence of gay-straight alliances in schools, and (4) school policies (nondiscrimination and antibullying) that specifically protected lesbian, gay, and bisexual students.
Lesbian, gay, and bisexual youth were significantly more likely to attempt suicide in the previous 12 months, compared with heterosexuals (21.5% vs 4.2%). Among lesbian, gay, and bisexual youth, the risk of attempting suicide was 20% greater in unsupportive environments compared to supportive environments. A more supportive social environment was significantly associated with fewer suicide attempts, controlling for sociodemographic variables and multiple risk factors for suicide attempts, including depressive symptoms, binge drinking, peer victimization, and physical abuse by an adult (odds ratio: 0.97 [95% confidence interval: 0.96–0.99]).
This study documents an association between an objective measure of the social environment and suicide attempts among lesbian, gay, and bisexual youth. The social environment appears to confer risk for suicide attempts over and above individual-level risk factors. These results have important implications for the development of policies and interventions to reduce sexual orientation–related disparities in suicide attempts.
Hatzenbuehler (2011) has presented interesting results by including community level variables in models predicting self-reported suicide attempts as a function of sexual orientation and other factors. While it was helpful that he included physical abuse by an adult as a predictor, the Oregon Healthy Teens Survey also included potentially useful variables such as physical abuse by a romantic partner during the past 12 months, sexual contact with an adult, and ever being forced to have sexual intercourse involuntarily, whose inclusion might have changed the model's outcomes with respect to the statistical significance of the community level factor. There were also community level variables, based on the student's perceptions of community attitudes towards teen smoking of cigarettes, use of marijuana, and regular use of alcoholic beverages, that could have supplemented the demographic aspects of community.
Sexual orientation was not only assessed by self-reported identity but also by self-reported sexual contact with males, females, or both males and females, a measure whose use might have changed the outcomes. Another community level variable that might be useful in future research to help explain suicide attempts might be fewer opportunities for same-sex sexual contact in certain (rural?) communities, which might make initial establishment of romantic relationships more difficult or re-establishment of the same after break-ups more difficult (with such break-ups a potential factor in depressed thinking).
There are other issues that merit further research. LGB orientation appears to be associated in some studies with higher rates of sexual activity [1] and earlier ages of onset of sexual activity [2]. Russell [3: 1253] reported how one GLB youth answered the question "What do you know about sexual minority youth?" by saying "We have all the fun!". Russell noted that "This statement beautifully illustrates the resilience that characterizes the lives of most sexual minority youth" [3: 1253]. It may also illustrate advantages of same-sex sexual orientation that may elicit jealousy, competition, or envy from heterosexual students, who may not care about sexual orientation per se but may resent its relative advantages in terms of variety or frequency of sexual contacts. However, the higher rates reported in this article for binge drinking among GLB students may suggest, among GLB students, lower levels of child/adolescent self-control, a factor for which higher levels have been found to predict better adult health, socioeconomic, and public safety outcomes in a longitudinal study from of children from age 3 to later adult age at 32 [4]. In some communities, acceptance of GLB identity may seem confounded with acceptance of binge drinking, drug abuse, or lower levels of self- control, making stigma against GLB youth seem to be a constructive way of promoting more mature levels of self-control among all youth, regardless of sexual orientation.
I also have some technical concerns with the article. On page 900 Figure 1 appears to indicate a maximum level of suicide intentions below 25% whereas the narrative indicates a high level of 25.47%. When I re- created the data set as best I could (there was about a 9% loss of data for community variables and the actual N's used in Figure 1 were not reported), it appeared that the effect size for suicide and sexual orientation was about 0.53, a medium-size effect [5]. The effect size overall for community environment appeared to be about 0.03, a very small effect, as Cohen [5] sets 0.20 or above as a small-size effect. Even among GLB youth, the effect size between community environment and suicide appeared to be on the order of 0.10 to 0.12, still far below Cohen's guidance for a small effect, as well as much lower than most of the other effects in the model. Even with a relatively large sample, the interaction effect portrayed in Figure 1 was reported as not significant statistically (page 899). It seems that such nonsignificance proved no barrier to the author reporting the results of the interaction in Figure 1 and then using Figure 1 to call for significant policy changes. I am not sure that's how science is normally done [6]. At the very least, the actual counts used to derive the percentages in Figure 1(25.47/20.37% for GLB youth and not specified precisely for heterosexual youth) should have been reported. A discussion of the advantages of using generalized estimating equations rather than hierarchical linear modeling would also have been useful for readers less familiar with the former.
[1] Udry JR, Chantala K. Risk assessment of adolescents with same- sex relationships. Journal of Adolescent Health 2002;31(1):84-92. [2] Saewyc EM, Bearinger LH, Heinz PA, Blum RM, Resnick MD. Gender differences in health and risk behaviors among bisexual and homosexual adolescents. Journal of Adolescent Health 1998;23(3): 181-188. [3] Russell ST. Sexual minority youth and suicide risk. The American Behavioral Scientist 2003;46(9):1241-1257. [4] Moffitt TE, Arseneault L, Belsky D, Dickson N, Hancox RJ, Harrington H, Houts R, Poulton R, Roberts BW, Ross S, Sears MR, Thomson WM, Caspi A. A gradient of childhood self-control predicts health, wealth, and public safety. PNAS 2011;108(7):2693-2698. [5] Cohen J. A power primer. Psychological Bulletin 1992;112(1):155-159. [6] Schumm WR. How science is done. Marriage and Family Review 2010;46(5):385-388.
Conflict of Interest:
Walter Schumm served as an expert witness for the state of Florida in a same-sex parent adoption trial in 2008.